BART-base is a large language model released by Meta, released under the Apache-2.0 license, with a claimed knowledge cutoff of 2019-10-29. Its weights, configuration, and tokenizer files have been cryptographically hashed and are queued for on-chain attestation by Weight Registry.
Description
Denoising sequence-to-sequence pretraining: corrupts text via token masking, deletion, infilling, sentence permutation, and document rotation, then reconstructs the original -- combining BERT's bidirectional encoder with GPT's autoregressive decoder. Particularly influential for abstractive summarization (CNN/DailyMail, XSum).
Metadata
Hugging Face stats
281,426 downloads on Hugging Face · 206 likes · pipeline: feature-extraction · library: transformers
Weight Hashes
Weights file
SHA-2561bd3ac8e5b3ac71c77cf18fbcd8b113e0bfceced94c5cfbbb9a2b4bb4781190b
SHA3-256a1229bae77c387378163e43f2db3264d146ddc38873578fa5d3c51e428cd0bca
BLAKE34d0a90ca1c2381a36626383079a1d9c0e7b2a2dc15c1a15fe8453febf9d67f9e
Config file
SHA-256c4022c931216afcb778991477332e5312aca8be478027ee3a8bf34b7004c44e2
SHA3-25658be51b3e6ff07134bffb9c1f6e3bf3a662ff7ea6eb1d61c72c389cc02edccf9
BLAKE393b841e2c428196a48d10996aae1858997ca57e1eea971c8796b3a74b0474405
Tokenizer file
SHA-256847bbeab6174d66a88898f729d52fa8d355fafe1bea101cf960dd404581df70e
SHA3-2560c697da2703e8ea995954b4a85cff8cbe531d8b9429cf74df59db238cb0f96ff
BLAKE34c2a3681edff22cad38aaaf3fcce153102d1e5146d55a39f0ac6632b8faf554e
Verify it yourself
curl -L -o "model.safetensors" "https://huggingface.co/facebook/bart-base/resolve/aadd2ab0ae0c8268c7c9693540e9904811f36177/model.safetensors" sha256sum "model.safetensors"
Expected SHA-256: 1bd3ac8e5b3ac71c77cf18fbcd8b113e0bfceced94c5cfbbb9a2b4bb4781190b